Abstract
Two approaches (Wolfensberger and Scull) to understanding services for people with mental disabilities are compared and seen to make complementary errors based upon the reduction of phenomena to ideology or economics respectively. Approaches combining both areas of analysis are seen as necessary in order to state what good services would look like and to understand the contexts in which services form and develop. It is argued that Wolfensberg's principle of normalisation allows a plausible statement of ideal criteria for service evaluation and design. A discussion follows of some possible constraints on its implementation under the present social order. A notion of 'prefigurative action research'is discussed as one way of pioneering more ideal services, while documenting the limits imposed on them at the present time.